I'd add one closing rule to step 9: a detect or mitigate action is done when someone has seen it work, not when the PR merges.
Your certificate alert is a good example. It can ship, pass review and still never page anyone, because it reads the wrong secret store or routes to a channel nobody watches. Issuing a 10-day certificate in staging and confirming the page reaches on-call is a short exercise, and it gives the action item something to link when it closes: the alert that fired.
That link pays off later too. When a future incident lists a missing alert as a contributing factor, and an older postmortem already has an action item for that alert, the closing evidence tells you whether it was tested or only merged.
I'd add one closing rule to step 9: a detect or mitigate action is done when someone has seen it work, not when the PR merges.
Your certificate alert is a good example. It can ship, pass review and still never page anyone, because it reads the wrong secret store or routes to a channel nobody watches. Issuing a 10-day certificate in staging and confirming the page reaches on-call is a short exercise, and it gives the action item something to link when it closes: the alert that fired.
That link pays off later too. When a future incident lists a missing alert as a contributing factor, and an older postmortem already has an action item for that alert, the closing evidence tells you whether it was tested or only merged.